基于感知哈希的图像内容鉴别性能分析  

Performance Analysis of Image Content Identification on Perceptual Hashing

作　　者：潘辉[1,2] 郑刚[1] 胡晓惠[1] 马恒太[1] 

机构地区：[1]中国科学院软件研究所天基综合信息系统重点实验室,北京100190 [2]中国科学院研究生院,北京100049

出　　处：《计算机辅助设计与图形学学报》2012年第7期925-931,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：中国科学院国防科技创新基金

摘　　要：感知哈希能有效地区分不同内容的图像,常用于检测盗版或重复的图像内容鉴别应用中.针对已有的对感知哈希的图像内容鉴别应用研究主要集中在算法设计上,缺少对其性能评价的理论方法的问题,提出从理论上评价内容鉴别性能的方法.该方法基于一类图像感知哈希算法的抗碰撞特性建立了内容鉴别的判定模型,并定义内容鉴别的性能公式,最后推导出统计学上的函数作为性能评价指标.实验结果表明,文中方法能很好地估计感知哈希在内容鉴别应用中的性能.Perceptual hashing can distinguish images of different contents effectively, and it is commonly used in content identification for detecting piracy or duplicate images. Published research of image content identification on perceptual hashing has mainly focused on the design of perceptual hashing algorithm, and very few theoretical methods have been developed for performance evaluation of content identification. Based on anti-collision property of image perceptual hashing, this paper introduces a mathematical model of content identification, defines performance formulas of content identification, and presents functions of the statistical distribution for performance evaluation. Experimental results show that the proposed approach estimates the performance of content identification on perceptual hashing very well.

关 键 词：感知哈希 图像内容鉴别 抗碰撞 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]